Нужно https://gist.github.com/Mihonarium/0ba242b419ebd2597a401632075bd634

voice_bot.php

<?php


include 'bot_functions.php';
function downloadAudioFile($audiolink)
{
	$ext = 'mp3';
	$saveFile = 'tmp/'.md5(microtime(true)).'.'.$ext; 
	file_put_contents("$saveFile", fopen("$audiolink", 'r'));
	return $saveFile;
}
function genAudioFile($mes, $lang, $speaker, $speachKitKey, $emotion, $speed)
{
	$genling = "https://tts.voicetech.yandex.net/generate?text=".urlencode($mes)."&format=mp3&lang={$lang}&speaker={$speaker}&key={$speachKitKey}&emotion={$emotion}&speed={$speed}";
	$genFile = downloadAudioFile($genling, $mes);
	return $genFile;
}

$confirmation_token = '';
$token = '';
$yandexVoiceToken = "";

$data = json_decode(file_get_contents('php://input'));
if(($data->type) == 'confirmation')
{
    echo $confirmation_token;
    return;
}
echo_ok();
if(($data->type) != 'message_new')
    return;

$user_id = $data->object->user_id;
vk_start_typing($user_id, $token);
$mes = $data->object->body;

message_send($user_id, '', $token, uploadVoiceMessage(genAudioFile($mes, 'ru', 'ermil', $yandexVoiceToken, 'neutral', '1.0'), $token, $user_id));